# find in which sublist of a list are any value of a list in python

I want to return a NumPy array of Element IDs with at least one Node being in the interesting Nodes list.

I have:

• A list of interesting Nodes, Interesting_Nodes.
The size is ~300,000 items.

• A list of sublists which represent Node forming an Element, Nodes.
The size is ~1,000,000 lists of ~20 values.

Nodes[i] contain the 20 Nodes creating Elements[i]. Which is an integer ID.

• A list of Elements, Elements. The size is ~1,000,000 items.

Here is an example of these lists.

import numpy as np
Interesting_Nodes=[1,2,10,40,400,1000]
Elements=[1,2,3]
Nodes=[[1,20,25],[30,400,35],[500,501,502]]


In this case the function will return [1,2] because Elements[0] have the Node 1 in the Interesting_Nodes list and Elements[1] contains the Node 400 which is also in the Interesting_Nodes list

I wrote this it seems to work but is very slow. Is there a way to perform this faster?

def recuperation_liste_element_dans_domaine_interet(Interesting_Nodes,Elements,Nodes):
Liste_elements=list([])
for n in Interesting_Nodes:
id=np.where(Nodes==n)
Liste_elements=Liste_elements+list(Elements[id[0]])
nbrListe = list(set(Liste_elements)) # remove all the element duplication
return np.array(nbrListe)


Another way to perform that I think could be: (be still too slow)

def recuperation_liste_element_dans_domaine_interet(Interesting_Nodes,Elements,Nodes):
Liste_elements=[0]*len(Elements)
compteur=0
j=0
for n in Nodes:
if any(i in Interesting_Nodes for i in n):
Liste_elements[compteur]=Elements[j]
compteur=compteur+1
j=j+1
Liste_elements=Liste_elements[:compteur]
return np.array(Liste_elements)


I'm assuming Elements is always a list of continuous indices from zero to the length of Nodes minus one and thus strictly not needed.

As a general guideline, explicit loops tend to be slow and also hard to read (and non-Pythonic).

To solve your problem efficiently, use data structures (namely sets):

import numpy as np

u = [1,2,10,40,400,1000]
n = [[1,20,25],[30,400,35],[500,501,502]]

# Convert them into sets
u = set(u)
n = [set(x) for x in n]

#  An individual bit is true if the corresponding list is "interesting"
idx = [bool(a.intersection(u)) for a in n]

# Perhaps convert to integer indices if you wish
np.where(idx)

• thanks a lot for your help. I figured out myself that set is much more powerful to do what I expected 1 hour ago ;) But not as "clean" as you write. I think it would do the job perfectly ! Thanks again ! Oct 2 '20 at 18:27
• By the way in fact i need Elements because it is not a continuous list. It coub be [10,1000,300]. But with idx I suppose I can juste write Elements[idx] to have my answer. Oct 2 '20 at 18:45
• Did you see an improvement in your "very slow" execution time? What kind of improvement did you see? Oct 3 '20 at 13:29
• a HUGE improvement !!! With the previous version, with the "utltimate" amount of data, i never manage to reach the end. With the new version, the code finish in few minutes. Thank you again @Juho Oct 4 '20 at 16:56